| Urban rainfall is known for its complicated mechanism and high spatial variability.Extreme rainfall events in urban area often cause serious flooding,which is detrimental to the functioning of cities.Therefore,as a primary way to improve the accuracy of flooding forecasts,adequate monitoring and forecasting of rainfall have potentially large implications for flood control and water resources exploitation and utilization in urban area.Based on multi-source rainfall observations,this study explores the small-scale variability of urban rainfall and develops a rainfall monitoring and forecasting system for the Greater Beijing Region(GBR).First,using the hourly rainfall observations from 118 gauges and a weather radar in the GBR during 2008~2012,the small-scale spatiotemporal characteristics of rainfall in this region was investigated.It is found that remarkable spatial heterogeneity exists in the diurnal patterns(temporal characteristics)of rainfall across the GBR.The diurnal cycles of urban and built-up area and its downwind area peak in the evening,whereas that of western mountainous area peaks in the afternoon.The geometrical features(spatial characteristics)of summer rainstorm was characterized using a rain-cell identification algorithm,and the results showed a highly localized pattern for summer GBR rainstorms,with typical scales of 4.31~20.58 km and 1.85~9.10 km in the major and minor radius directions,respectively.Given that the representative area per gauge of current gauge network is ~140 km2,the current gauge network is insufficient to capture localized rainstorms during summer.Given the deficiency of gauge measurement,an X-band weather radar has been employed to derive high spatial resolution data.This study first developed the radar quantitative precipitation estimation(QPE)algorithms for the GBR.The radar QPE algorithm contains 7 steps,namely radar calibration,non-precipitating echo removal,attenuation correction,beam integration,convective–stratiform classification,multiple ZR translation and resampling.Analysis of the results show that,although the radar QPE underestimates rainfall at the hourly scale compared with gauges,it can provide promising rainfall data after all the above corrections.The study highlights the necessity of deriving localized Z-R relationships for different rainfall types and conducting volumetric reflectivity measurements to do the beam integration.Moreover,an advection method is used to correct the temporal sampling error,and it is found that the advection correction tends to enhance the radar QPE accuracy for fast-moving intense rain events.Finally,an ensemble nowcasting system for GBR was developed,including two primary nowcasting methods.The first method,named as pixel-based nowcasting(PBN),relies on extrapolation of radar images,while the second method is a storm-scale numerical weather prediction(NWP)model.The 3DVAR and Cloud Analysis method is used to assimilate radar observed reflectivity and radial velocity data which can provide an accurate initial condition for the NWP model.Furthermore,PBN and NWP were merged by implementing a weighted average of the single nowcasting product.The results show that PBN method is effective in tracking and nowcasting rainfall events in an hour.It performs better in nowcasting stratiform events than convective events.Meanwhile,storm-scale NWP that assimilated reflectivity and radial velocity data can provide a promising result in an hour.However,merging PBN with storm-scale NWP does not always provide better results.It is therefore recommended that the system should provide multiple nowcasting results rather than a single merge result. |
PDF Full Text Request